Merge artifact splitting work

Rationale
=========

This patch series implements the concept of stratum splitting.

For a long time we've had code to split a chunk into multiple artifacts,
however there's not been a way to split strata up, or systems select a
subset of the produced stratum artifacts to be included in the system.

This patch series implements the ability to split strata and have systems
include them in a way which still has the same behaviour if no rules
are specified, but with default rules that split chunk artifacts up into
various components, strata into runtime and development versions and has
systems include everything by default, but can be told to include less.

The default rules have chunk foo split up into -bins, -libs, -devel,
-doc, -locale and -misc.

These rules can be overridden in the chunk morphology by adding the new
'products' field, which lists match rules like the following:

    products:
    - artifact: libudev
      include:
      - (usr/)?lib(32|64)?/lubg?udev\..*
    - artifact: udev
      include:
      - (usr/)?s?bin/udev*
      - (usr/)?lib(32|64|exec)?/systemd/systemd-udevd

Strata are by default split into -runtime and -devel. -devel by default
contains chunks ending with -devel and -doc, -runtime contains everything
else.

Extra match rules can be added to a stratum similarly to chunks, but
instead of matching file names, they match artifact names.

    products:
    - artifact: core-python
      include:
      - "cpython-.*" # lazy shortcut to put all of cpython in this stratum
      - "python-.*" # lazy shortcut to include all python chunks in

Additionally, in chunk specs, chunk artifacts may be assigned to stratum
artifacts, this takes precedence over products match rules in the stratum
and the default match rules. Assigning the chunk to `null` will discard
the chunk.

    chunks:
    ...
    - name: systemd
      ...
      artifacts:
          libudev: foundation-runtime
          udev: foundation-runtime
          systemd-doc: null

By default a system includes every produced artifact of every stratum
listed. Instead a subset can be specified in the stratum spec as follows:

    name: tiny-system
    strata:
    - name: build-essential
      ...
      artifacts:
      - build-essential-runtime

1 files changed, 303 insertions, 0 deletions

diff --git a/morphlib/artifactsplitrule.py b/morphlib/artifactsplitrule.py
new file mode 100644
index 00000000..246691d8
--- /dev/null
+++ b/morphlib/artifactsplitrule.py
@@ -0,0 +1,303 @@
+# Copyright (C) 2013-2014  Codethink Limited
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; version 2 of the License.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License along
+# with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+
+import collections
+import itertools
+import re
+
+import morphlib
+
+
+class Rule(object):
+    '''Rule base class.
+
+    Rules are passed an object and are expected to determine whether
+    it matches. It's roughly the same machinery for matching files
+    as artifacts, it's just that Files are given just the path, while
+    Artifact matches are given the artifact name and the name of the
+    source it came from.
+
+    '''
+
+    def match(self, *args):
+        return True
+
+
+class FileMatch(Rule):
+    '''Match a file path against a list of regular expressions.
+
+    If the path matches any of the regular expressions, then the file
+    is counted as a valid match.
+
+    '''
+
+    def __init__(self, regexes):
+        # Possible optimisation: compile regexes as one pattern
+        self._regexes = [re.compile(r) for r in regexes]
+
+    def match(self, path):
+        return any(r.match(path) for r in self._regexes)
+
+
+class ArtifactMatch(Rule):
+    '''Match an artifact's name against a list of regular expressions.
+    '''
+
+    def __init__(self, regexes):
+        # Possible optimisation: compile regexes as one pattern
+        self._regexes = [re.compile(r) for r in regexes]
+
+    def match(self, (source_name, artifact_name)):
+        return any(r.match(artifact_name) for r in self._regexes)
+
+
+class ArtifactAssign(Rule):
+    '''Match only artifacts with the specified source and artifact names.
+
+    This is a valid match if the source and artifact names exactly match.
+    This is used for explicit artifact assignment e.g. chunk artifact
+    foo-bins which comes from chunk source foo goes into stratum
+    bar-runtime.
+
+    '''
+
+    def __init__(self, source_name, artifact_name):
+        self._key = (source_name, artifact_name)
+
+    def match(self, (source_name, artifact_name)):
+        return (source_name, artifact_name) == self._key
+
+
+class SourceAssign(Rule):
+    '''Match only artifacts which come from the specified source.
+
+    This is a valid match only if the artifact comes from the specified
+    source. e.g. all artifacts produced by source bar-runtime go into
+    system baz
+
+    '''
+
+    def __init__(self, source_name):
+        self._source = source_name
+
+    def match(self, (source_name, artifact_name)):
+        return source_name == self._source
+
+
+class SplitRules(collections.Iterable):
+    '''Rules engine for splitting a source's artifacts.
+
+    Rules are added with the .add(artifact, rule) method, though another
+    SplitRules may be created by passing a SplitRules to the constructor.
+
+    .match(path|(source, artifact)) and .partition(iterable) are used
+    to determine if an artifact matches the rules. Rules are processed
+    in order, so more specific matches first can be followed by more
+    generic catch-all matches.
+
+    '''
+
+    def __init__(self, *args):
+        self._rules = list(*args)
+
+    def __iter__(self):
+        return iter(self._rules)
+
+    def add(self, artifact, rule):
+        self._rules.append((artifact, rule))
+
+    @property
+    def artifacts(self):
+        '''Get names of all artifacts in the rule set.
+
+        Returns artifact names in the order they were added to the rules,
+        and not repeating the artifact.
+
+        '''
+
+        seen = set()
+        result = []
+        for artifact_name, rule in self._rules:
+            if artifact_name not in seen:
+                seen.add(artifact_name)
+                result.append(artifact_name)
+        return result
+
+    def match(self, *args):
+        '''Return all artifact names the given argument matches.
+
+        It's returned in match order as a list, so it's possible to
+        detect overlapping matches, even though most of the time, the
+        only used entry will be the first.
+
+        '''
+
+        return [a for a, r in self._rules if r.match(*args)]
+
+    def partition(self, iterable):
+        '''Match many files or artifacts.
+
+        It's the common case to take a bunch of filenames and determine
+        which artifact each should go to, so rather than implement this
+        logic in multiple places, it's here as a convenience method.
+
+        '''
+
+        matches = collections.defaultdict(list)
+        overlaps = collections.defaultdict(set)
+        unmatched = set()
+
+        for arg in iterable:
+            matched = self.match(arg)
+            if len(matched) == 0:
+                unmatched.add(arg)
+                continue
+            if len(matched) != 1:
+                overlaps[arg].update(matched)
+            matches[matched[0]].append(arg)
+
+        return matches, overlaps, unmatched
+
+
+# TODO: Work out a good way to feed new defaults in. This is good for
+#       the usual Linux userspace, but we may find issues and need a
+#       migration path to a more useful set, or develop a system with
+#       a different layout, like Android.
+DEFAULT_CHUNK_RULES = [
+    ('-bins', [ r"(usr/)?s?bin/.*" ]),
+    ('-libs', [
+        r"(usr/)?lib(32|64)?/lib[^/]*\.so(\.\d+)*",
+        r"(usr/)libexec/.*"]),
+    ('-devel', [
+        r"(usr/)?include/.*",
+        r"(usr/)?lib(32|64)?/lib.*\.a",
+        r"(usr/)?lib(32|64)?/lib.*\.la",
+        r"(usr/)?(lib(32|64)?|share)/pkgconfig/.*\.pc"]),
+    ('-doc', [
+        r"(usr/)?share/doc/.*",
+        r"(usr/)?share/man/.*",
+        r"(usr/)?share/info/.*"]),
+    ('-locale', [
+        r"(usr/)?share/locale/.*",
+        r"(usr/)?share/i18n/.*",
+        r"(usr/)?share/zoneinfo/.*"]),
+    ('-misc', [ r".*" ]),
+]
+
+
+DEFAULT_STRATUM_RULES = [
+    ('-devel', [
+        r'.*-devel',
+        r'.*-debug',
+        r'.*-doc']),
+    ('-runtime', [
+        r'.*-bins',
+        r'.*-libs',
+        r'.*-locale',
+        r'.*-misc',
+        r'.*']),
+]
+
+
+def unify_chunk_matches(morphology):
+    '''Create split rules including defaults and per-chunk rules.
+
+    With rules specified in the morphology's 'products' field and the
+    default rules for chunks, generate rules to match the files produced
+    by building the chunk to the chunk artifact they should be put in.
+
+    '''
+
+    split_rules = SplitRules()
+
+    for ca_name, patterns in ((d['artifact'], d['include'])
+                              for d in morphology['products']):
+        split_rules.add(ca_name, FileMatch(patterns))
+
+    name = morphology['name']
+    for suffix, patterns in DEFAULT_CHUNK_RULES:
+        ca_name = name + suffix
+        # Default rules are replaced by explicit ones
+        if ca_name in split_rules.artifacts:
+            break
+        split_rules.add(ca_name, FileMatch(patterns))
+
+    return split_rules
+
+
+def unify_stratum_matches(morphology):
+    '''Create split rules including defaults and per-stratum rules.
+
+    With rules specified in the chunk spec's 'artifacts' fields, the
+    stratum's 'products' field and the default rules for strata, generate
+    rules to match the artifacts produced by building the chunks in the
+    strata to the stratum artifact they should be put in.
+
+    '''
+
+    assignment_split_rules = SplitRules()
+    for spec in morphology['chunks']:
+        source_name = spec['name']
+        for ca_name, sta_name in sorted(spec.get('artifacts', {}).iteritems()):
+            assignment_split_rules.add(sta_name,
+                                       ArtifactAssign(source_name, ca_name))
+
+    # Construct match rules separately, so we can use the SplitRules object's
+    # own knowledge of which rules already exist to determine whether
+    # to include the default rule.
+    # Rather than use the existing SplitRules, use a new one, since
+    # match rules suppliment assignment rules, rather than replace.
+    match_split_rules = SplitRules()
+    for sta_name, patterns in ((d['artifact'], d['include'])
+                               for d in morphology.get('products', {})):
+        match_split_rules.add(sta_name, ArtifactMatch(patterns))
+
+    for suffix, patterns in DEFAULT_STRATUM_RULES:
+        sta_name = morphology['name'] + suffix
+        if sta_name in match_split_rules.artifacts:
+            break
+        match_split_rules.add(sta_name, ArtifactMatch(patterns))
+
+    # Construct a new SplitRules with the assignments before matches
+    return SplitRules(itertools.chain(assignment_split_rules,
+                                      match_split_rules))
+
+
+def unify_system_matches(morphology):
+    '''Create split rules including defaults and per-chunk rules.
+
+    With rules specified in the morphology's 'products' field and the
+    default rules for chunks, generate rules to match the files produced
+    by building the chunk to the chunk artifact they should be put in.
+
+    '''
+
+    name = morphology['name'] + '-rootfs'
+    split_rules = SplitRules()
+
+    for spec in morphology['strata']:
+        source_name = spec.get('name', spec['morph'])
+        if spec.get('artifacts', None) is None:
+            split_rules.add(name, SourceAssign(source_name))
+            continue
+        for sta_name in spec['artifacts']:
+            split_rules.add(name, ArtifactAssign(source_name, sta_name))
+
+    return split_rules
+
+
+def unify_cluster_matches(_):
+    return None